ISPRS Istanbul Workshop 2010 on Modeling of optical airborne and spaceborne Sensors, WG I/4, Oct. 11-13, IAPRS Vol. XXXVIII-1/W17.
MECHANICAL FMC TECHNIQUES FOR REMOVING IMAGE BLUR IN DIGITAL
AERIAL CAMERAS
O. Selimoglu, O. Yilmaz, O. Şengül, and B. Akarca
TUBITAK Uzay ODTU Kampüsü 06531 Ankara
ozgur.selimoglu,ozgur.yilmaz,orhan.sengul,beyza.akarca@uzay.tubitak.gov.tr
KEY WORDS: FMC, Image Blur, Piezo-actuators, Mechanical FMC
ABSTRACT:
Image blur caused by high speed motion is one of the main problems of aerial digital cameras. Removing image blur with special
techniques is called Forward Motion Compensation (FMC) techniques. During the exposure time of the sensor, image of the same
earth region moves more than many pixels on the sensor due to high speed, and causes image blur. Expensive TDI (time delay
integration) sensors are now available and electronically compensate the motion within the sensor. In this article we propose
mechanical compensation methods. During the exposure time, movement of the camera with proper angular speed will result in
FMC. This can be achieved by pure older mechanical methods. Another method is to move the sensor linearly, proportional the
speed of image walk. This can be provided by piezo-actuators. In this paper we present examples of these two mechanical FMC
techniques. By these methods cheaper standard sensors can be used in digital aerial cameras.
1. INTRODUCTION
FMC is a necessary requirement for most of the aerial camera
systems. There are electronic, mechanical and software methods
to achieve FMC. Electronic FMC needs sophisticated detector
structure and most of the time it is difficult to obtain these
sensors. Instead of electronic FMC, Mechanical techniques can
be used to achieve FMC without sophisticated equipment.
V
W
Angular movement of camera system and just moving the
sensor head are two methods for mechanical FMC. The details
of these two methods are given in the following paragraphs.
θ
h
2. ANGULAR MOVEMENT OF CAMERA SYSTEM:
As the aerial platform moves while taking the pictures of the
earth, relative motion at exposure period causes image blur. We
can analyse the situation by defining a realistic scenario.
In the scenario shown in Figure -1, the plane moves on the sky
with a speed V (125 knot = 231.5 km/hr) with respect to the
ground while pointing the nadir. The flying altitude is taken as
1500 m and the camera is configured to get a ground sampling
distance of 15 cm with an exposure time t (5 millisecond).
Anybody experiences to see the nearby trees while travelling in
a car that, you have to move your head towards the opposite
direction of the travel. The same concept can be a solution for
the image blur in remote sensing applications.
s
Figure -1: Plane moving while taking pictures of the earth.
In the plane, we can turn the camera head with a constant speed
to remove image blur. This angular speed is not so wild for
most of the civilian applications and can be achieved by using
simple electro-mechanical systems.
For the nadir point, we can calculate the necessary rotational
speed by equation 1 to remove the image blur. We only need
the velocity of the aircraft and the flying altitude.
ω=
v
h
(1)
ISPRS Istanbul Workshop 2010 on Modeling of optical airborne and spaceborne Sensors, WG I/4, Oct. 11-13, IAPRS Vol. XXXVIII-1/W17.
If we want FMC for all the image points, we need to calculate
the necessary rotational speed for those corresponding points.
As a result of calculations we conclude that different rotational
speeds are necessary for different points on the ground. The
necessary rotational speed is calculated by using the equation 2.
ω=
θ − arctan(tan θ −
t
vt
)
h
(2)
The necessary angular speed needed for different angular
position is shown in the Figure-2. As shown from the graph,
necessary angular velocity reduces with the angle.
Figure -3: Remaining image blur with respect to the angle from
nadir.
From the mechanical point of view, moving all the camera
structure can induce vibration problems. So, if the camera
structure is not rigid enough, instead of moving the entire
camera, a mirror can be added to the system. The camera will
look to the mirror and see the ground from this mirror. If we
rotate the mirror, it will have the same effect of rotating the
camera with respect to the ground.
3. LINEAR MOVEMENT OF THE SENSOR:
Figure -2: Required rotational speed for FMC
Harmonic motion is needed to get prepared to the next shot.
The exposure time is a very small fraction of the time between
two shots. In the remaining time, the camera should return to
the initial position again and at the next exposure time it should
have the proper angular speed. This movement is easily can
achieved by using a constant rotation speed cam assembly. By
giving proper shape to the cam, we can easily rotate the camera
body with a proper speed profile.
An inherent behaviour of this kind of FMC is that a complete
FMC is not possible. Even if we can achieve a complete FMC
for the nadir point, for the other ground points it is not possible.
But the remaining image blur (smearing) may be acceptable
depending on the FOV of the objective and the necessary
compensation for the application. In the Figure-3, we have
shown that which angular deviation from nadir point has which
amount of image blur.
From simulations we conclude that this technique gives
satisfactory results for small FOV systems. If we put a criteria
such that image blur should not exceed half of a pixel, than the
objective that we will be used should not have a full field of
view greater than 60 degrees. Absolute FMC is achieved at the
nadir and FMC error increases with the angular deviation.
Without moving all the camera structure, it is possible to move
only the sensor part linearly towards the direction of image
walk. In this method, only a small linear movement is sufficient
to remove image blur.
To get FMC, during exposure, a constant speed linear
movement is necessary. The required motion can be obtained
by using small linear displacement actuators such as piezo
actuators. To achieve a uniform movement, it is better to start
the movement a little earlier than the exposure. Necessary focal
plane movement speed can be calculated by Equation-3.
V fp = v *
f
h
(3)
If we consider the same scenario used in the first method, the
only necessary variable to calculate speed of the focal plane
movement is the focal length. By selecting the focal length as
70 mm which is an acceptable focal length, than the required
speed is only 3 mm/sec.
The movement should start a little earlier than the exposure and
finish a little time later. If we assume the movement starts an
exposure duration earlier and finish an exposure duration later,
than the actuator should move totally 3 exposure durations. If
we choose the exposure time 5 millisecond, than only 45
micron movement range is necessary. This is in the range of
simple piezo actuators.
ISPRS Istanbul Workshop 2010 on Modeling of optical airborne and spaceborne Sensors, WG I/4, Oct. 11-13, IAPRS Vol. XXXVIII-1/W17.
The same movement should be repeated for every shot. To get
prepared to the next shot, the sensor should move back to the
original position after the exposure time.
In this configuration, the same focal plane movement speed is
necessary for nadir and off axis points. As a result, a complete
FMC can be achieved.
4. PROBLEMS
PERSPECTIVE
CAUSED
BY
DISTORTION
AND
Mathematical models assume perfect optics and perfect nadir
pointing. But the situation is not always like this. Optics has
always some distortion. This optical distortion may prevent
FMC in the aerial photography.
Distortion is the changing effective focal length with the angle.
As seen from the Equation-3, focal length is a parameter for the
FMC generating movement. Therefore if the distortion is large,
a successful FMC is not possible by moving the focal plane
(method 2). The first method will not be affected from the
distortion because it is not related to the focal length.
In some applications objective lenses should be placed with an
angle from the vertical. The most encountered application is
large FOV imaging. To achieve large FOV, multiple lenses
should be accommodated in the camera. Every lens will look
different directions and cover different land areas. This will
cause FMC degradation because objective lenses are optimised
for nadir pointing photography. If the camera look with an
oblique angle, than there are perspective problems. Since all the
pixels do not correspond to the same ground distance coverage,
it is not possible to achieve a complete FMC by using simple
methods.
5. CONCLUSION
In this paper two different mechanical FMC techniques are
discussed. Both of the methods have inherent strong and weak
sides.
For multiple objective systems, single cam actuator is enough if
we use the first method. By moving the actuator, all the
cameras in the camera system will experience image blur
reduction.
If we use second method with multiple objective systems, than
every detector should have its own actuator. While the second
method is more successful to achieve FMC, the first method is
simpler.
In the high performance expensive camera systems, multiple
focal plane actuators can be used where the price is not the
main parameter. For low cost systems, adding FMC by using
first technique is easy and will significantly reduce the image
blur to an acceptable level.
6. REFERENCES
1.
Sandau, R. (2010). Digital Airborne Camera,
Introduction and Technology. Springer.
2.
Recon/optical Inc. Electro-optical imaging array with
motion compensation. 1991.(PCT/US1991/008649)